MNRAS 450, 288–294 (2015) doi:10.1093/mnras/stv488 Detection of a supervoid aligned with the cold spot of the cosmic microwave background Istvan´ Szapudi,1‹ Andras´ Kovacs,´ 2,3,4 Benjamin R. Granett,5 Zsolt Frei,2,3 Joseph Silk,6 Will Burgett,1 Shaun Cole,7 Peter W. Draper,7 Daniel J. Farrow,7 Nicholas Kaiser,1 Eugene A. Magnier,1 Nigel Metcalfe,7 Jeffrey S. Morgan,1 Paul Price,8 John Tonry1 and Richard Wainscoat1 1Institute for Astronomy, University of Hawaii 2680 Woodlawn Drive, Honolulu, HI 96822, USA 2Institute of Physics, Eotv¨ os¨ Lorand´ University, Pazm´ any´ Peter´ set´ any´ 1/A, 1117 Budapest, Hungary 3MTA-ELTE EIRSA ‘Lendulet’¨ Astrophysics Research Group, Pazm´ any´ Peter´ set´ any´ 1/A, 1117 Budapest, Hungary 4Institut de F´ısica d’Altes Energies, Universitat Autonoma´ de Barcelona, E-08193 Bellaterra (Barcelona), Spain 5INAF OA Brera, Via E. Bianchi 46, I-23807 Merate, Italy 6Department of Physics and Astronomy, The Johns Hopkins University, Baltimore, MD 21218, USA 7Department of Physics, Durham University, South Road, Durham DH1 3LE, UK 8Department of Astrophysical Sciences, Princeton University, Princeton, NJ 08544, USA Accepted 2015 March 4. Received 2015 February 24; in original form 2014 May 6 ABSTRACT We use the WISE-2MASS infrared galaxy catalogue matched with Pan-STARRS1 (PS1) galaxies to search for a supervoid in the direction of the cosmic microwave background (CMB) cold spot (CS). Our imaging catalogue has median redshift z 0.14, and we obtain photometric redshifts from PS1 optical colours to create a tomographic map of the galaxy distribution. The radial profile centred on the CS shows a large low-density region, extending over tens of degrees. Motivated by previous CMB results, we test for underdensities within two angular radii, 5◦, and 15◦. The counts in photometric redshift bins show significantly low densities at high detection significance, 5σ and 6σ, respectively, for the two fiducial radii. The line-of-sight position of the deepest region of the void is z 0.15–0.25. Our data, combined with an earlier measurement by Granett, Szapudi & Neyrinck, are consistent with a −1 large Rvoid = (220 ± 50) h Mpc supervoid with δm −0.14 ± 0.04 centred at z = 0.22 ± 0.03. Such a supervoid, constituting at least a 3.3σ fluctuation in a Gaussian distribution of the cold dark matter model, is a plausible cause for the CS. Key words: surveys – cosmic background radiation – cosmology: observations – large-scale structure of Universe. CS (Inoue & Silk 2006, 2007; Inoue, Sakai & Tomita 2010). The 1 INTRODUCTION latter would be readily detectable in large-scale structure surveys The cold spot (CS) of the cosmic microwave background (CMB) thus motivating several observational studies. is an exceptionally cold −70 µK area centred on (l, b) (209◦, A low-density region approximately aligned with the CS was de- −57◦) Galactic coordinates. It was first detected in the Wilkinson tected in a catalogue of radio galaxies (Rudnick, Brown & Williams Microwave Anisotropy Probe (Bennett et al. 2013)mapsat 3σ 2007), although its significance has been disputed (Smith & Huterer significance using wavelet filtering (Vielva et al. 2004;Cruzetal. 2010). A targeted redshift survey in the area (Bremer et al. 2010) 2005). The CS is perhaps the most significant among the ‘anoma- found no evidence for a void in the redshift range of 0.35 <z<1, lies’, potential departures from isotropic and/or Gaussian statistics, while an imaging survey with photometric redshifts (Granett, and all confirmed by Planck (Planck Collaboration XXIII 2014). Szapudi & Neyrinck 2010) excluded the presence of a large un- Explanations of the CS range from statistical fluke through hitherto derdensity of δ −0.3 between redshifts of 0.5 <z<0.9 and undiscovered physics, e.g. textures (Cruz et al. 2008; Vielva 2010), finding none at 0.3 <z<0.5. Both of these surveys ran out of to the linear and non-linear ISW effect (Sachs & Wolfe 1967; Rees volume at low redshifts due to their small survey area, although & Sciama 1968) from a 200 h−1 Mpc supervoid centred on the the data are consistent with the presence of a void at z<0.3 with low significance (Granett et al. 2010). In a shallow photometric red- shift catalogue constructed from the Two Micron All Sky Survey E-mail: [email protected] (2MASS; Skrutskie et al. 2006) and SuperCOSMOS (Hambly et al. C 2015 The Authors Published by Oxford University Press on behalf of the Royal Astronomical Society PS1-WISE-2MASS void 289 Figure 1. The left-hand panel shows the photo-z accuracy achieved by the SVM. Dotted lines indicate the σ z ≈ 0.034 1σ error bars around the expectation. The right-hand panel illustrates the normalized redshift distributions of our subsamples used in the photo-z pipeline: training and control sets selected in GAMA, photo-z distributions estimated for the WISE-2MASS-PS1-GAMA control sample, and photo-zs of interest in the WISE-2MASS-PS1 matched area. The median redshift of all subsamples is z 0.14. 2001) with a median redshift of z = 0.08 an underdensity was found W1WISE ≤ 15.2 mag and W1WISE − J2MASS ≤−1.7. We add a (Francis & Peacock 2010) that can account for a CMB decrement further limit of J2MASS ≤ 16.5 mag to ensure spatial homogeneity of T −7 µK in the standard cold dark matter (CDM) cos- based on our experiments. This refinement shifts the median red- mology. While so far no void was found that could fully explain the shift of the sample to z 0.14. The catalogue covers 21 200 deg2 CS, there is strong, 4.4σ , statistical evidence that superstructures after masking. We mask pixels with E(B − V) ≥ 0.1, and regions at imprint on the CMB as cold and hot spots (Granett, Neyrinck & galactic latitudes |b| < 20◦ to exclude potentially contaminated re- Szapudi 2008, 2009;Papai´ & Szapudi 2010; Cai et al. 2014; Planck gions near the Galactic plane (Schlegel, Finkbeiner & Davis 1998). Collaboration XXIII 2014). Note that the imprinted temperature in These conservative limits result in a data set deeper than the 2MASS all of these studies is significantly colder than simple estimates from Extended Source Catalog (Jarrett et al. 2000) and more uniform than linear ISW (e.g. Rudnick et al. 2007;Papai´ & Szapudi 2010;Papai,´ WISE (Kovacs´ et al. 2013). These galaxies have been matched with Szapudi & Granett 2011) would suggest. PS1 objects within a 50◦ × 50◦ area centred on the CS, except for a The Wide-field Infrared Survey Explorer (WISE; Wright et al. Dec. ≥−28.0 cut to conform to the PS1 boundary. We used PV1.2 2010) all-sky survey effectively probes low redshift z ≤ 0.3 uncon- reprocessing of PS1 in an area of 1300 deg2. strained by previous studies. Using the WISE-2MASS all sky galaxy For matching we applied a nearest neighbour search using the map of Kovacs´ & Szapudi (2015) as a base catalogue, we match SCIPY kd-TREE algorithm with 1-arcsec matching radius, finding a a 1300 deg2 area with the PV1.2 reprocessing of Pan-STARRS1 PS1 pair for 86 per cent of the infrared galaxies, and resulting 73 100 (hereafter PS1; Kaiser 2004), adding optical colours for each ob- objects in the final catalogue. Galaxies without a PS1 match are faint ject. In the resulting catalogue with photometric redshifts we test in the optical, and predominantly massive early-type galaxies at for the presence of a large low-density region, a supervoid, cen- z>1(Yanetal.2013). For PS1, we required a proper measurement tred on the CS. We defined the centre of the CS from the latest of Kron (Kron 1980) and PSF magnitudes in gP1, rP1,andiP1 bands Planck results (Planck Collaboration XXIV 2014). Based on the that were used to construct photometric redshifts (photo-zs) with a ◦ literature, we decided in advance to test for an underdensity at 5 Support Vector Machine (SVM) algorithm, and the PYTHON SCIKIT- (Vielva et al. 2004;Cruzetal.2005; Rudnick et al. 2007;Bremer LEARN (Pedregosa et al. 2011) routines in regression mode. The et al. 2010; Granett et al. 2010) and 15◦ (Inoue et al. 2010; Zhang & training and control sets were created matching WISE-2MASS, Huterer 2010) of radii. The fact that these values gleaned from CMB PS1, and the Galaxy and Mass Assembly (GAMA; Driver et al. independently of our (large-scale structure) data simplifies the in- 2011) redshift survey. We chose a Gaussian kernel for our SVM terpretation of our results in the Bayesian framework, in particular, and trained on 80 per cent of the GAMA redshifts, while the rest minimize any a posteriori bias. were used for a control set. We empirically tuned the standard SVM The paper is organized as follows. Data sets and map-making parameters, finding the best performance when using C = 10.0, algorithms are described in Section 2; our observational results and γ= 0.1. We characterize our photo-z quality with the error 2 are presented in Section 3; the final section contains a summary, σz = (zphot − zspec) , finding σ z ≈ 0.034, as summarized in discussion, and interpretation of our results. Fig. 1. 2 DATA SETS AND METHODOLOGY 3 RESULTS Initially, we select galaxies from the WISE-2MASS catalogue The projected WISE-2MASS galaxy density field along with (Kovacs´ & Szapudi 2015) containing sources to flux limits of the Planck Spectral Matching Independent Component Analysis MNRAS 450, 288–294 (2015) 290 I. Szapudi et al. Figure 2. Gnomonic projections of the WISE-2MASS projected density map (left) and the Planck SMICA CMB map (right).